%reload_ext autotime
import pandas as pd
import requests
from pprint import pprint
import json
import torch
from PIL import Image
from transformers import MllamaForConditionalGeneration, AutoProcessor
from tqdm.auto import tqdm

pd.options.plotting.backend = "plotly"
pd.set_option("display.max_columns", None)
pd.set_option("display.max_colwidth", 100)

✔️ 4.72 s (2024-12-12T09:23:59/2024-12-12T09:24:03)

2024-12-12 09:24:02.300884: I tensorflow/core/util/port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
2024-12-12 09:24:02.313220: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:485] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
2024-12-12 09:24:02.329137: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:8454] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
2024-12-12 09:24:02.334326: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1452] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
2024-12-12 09:24:02.348072: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2024-12-12 09:24:03.164700: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT

df = pd.read_csv("results.csv").drop_duplicates(subset="panoid")
df

✔️ 20.4 ms (2024-12-12T09:45:14/2024-12-12T09:45:14)

# Loading this model needs about 22.69GB of GPU memory
model_id = "meta-llama/Llama-3.2-11B-Vision-Instruct"

model = MllamaForConditionalGeneration.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)
processor = AutoProcessor.from_pretrained(model_id)

✔️ 13.6 s (2024-12-12T09:24:04/2024-12-12T09:24:17)

The model weights are not tied. Please use the `tie_weights` method before using the `infer_auto_device` function.

Loading checkpoint shards:   0%|          | 0/5 [00:00<?, ?it/s]

for row in tqdm(df.sample(10).itertuples(index=False)):
    panoid = row.panoid
    image = Image.open(f"panoramas/{panoid}.jpg")
    display(image.resize((image.width // 2, image.height // 2)))
    messages = [
        {
            "role": "user",
            "content": [
                {"type": "text", "text": """
                    This image is a panorama from Google Street View.
                    From the image, extract the following information, in JSON format:
                    green: Percentage of the image that is green space (e.g. parks, gardens, trees, grass etc.). A number from 0-100.
                    environment: Classify the nature of the environment in this image. Built up/green/residential/shops/cafes?. A string.
                    water: If you see any streams/ponds/rivers/ocean in the image, estimate the distance to the water in meters. A number. If there is no water, return 0.
                    obscured: Proportion of view obscured by buildings (how much of total line of sight is blocked by buildings in close proximity). A number from 0-100.
                    people: the number of people you see in the image
                    cars: the number of cars you see in the image
                    bikes: the number of bikes you see in the image

                    Do not include comments in your JSON response. Only respond with the JSON object. Make sure the JSON is valid.
                """},
                {"type": "image"},
            ]
        }
    ]
    input_text = processor.apply_chat_template(messages, add_generation_prompt=True)
    inputs = processor(
        image,
        input_text,
        add_special_tokens=False,
        return_tensors="pt"
    ).to(model.device)

    for retry in range(3):
        output = model.generate(**inputs, max_new_tokens=5000)
        result = processor.decode(output[0])
        result = result[result.rindex("<|end_header_id|>") + len("<|end_header_id|>"):].strip().replace("<|eot_id|>", "")
        print("Output:")
        try:
            result = json.loads(result)
            pprint(result)
            print("\n")
            break
        except json.JSONDecodeError:
            print(f"Unable to parse: {result}")

✔️ 26.5 s (2024-12-12T09:51:22/2024-12-12T09:51:49)

0it [00:00, ?it/s]

Output:
{'bikes': 0,
 'cars': 0,
 'environment': 'built up',
 'green': 0,
 'obscured': 100,
 'people': 0,
 'water': 0}

Output:
{'bikes': 0,
 'cars': 0,
 'environment': 'shops',
 'green': 0,
 'obscured': 100,
 'people': 0,
 'water': 0}

Output:
{'bikes': 0,
 'cars': 0,
 'environment': 'built up',
 'green': 0,
 'obscured': 100,
 'people': 0,
 'water': 0}

Output:
{'bikes': 0,
 'cars': 0,
 'environment': 'built up',
 'green': 0,
 'obscured': 50,
 'people': 0,
 'water': 0}

Output:
{'bikes': 0,
 'cars': 0,
 'environment': 'Built up',
 'green': 0,
 'obscured': 100,
 'people': 0,
 'water': 0}

Output:
{'bikes': 0,
 'cars': 0,
 'environment': 'shops',
 'green': 0,
 'obscured': 0,
 'people': 0,
 'water': 0}

Output:
{'bikes': 0,
 'cars': 0,
 'environment': 'park',
 'green': 65,
 'obscured': 20,
 'people': 0,
 'water': 0}

Output:
{'bikes': 0,
 'cars': 0,
 'environment': 'residential',
 'green': 50,
 'obscured': 50,
 'people': 0,
 'water': 0}

results = []
for row in tqdm(df.itertuples(index=False), total=len(df)):
    panoid = row.panoid
    image = Image.open(f"panoramas/{panoid}.png")
    #display(image)
    messages = [
        {
            "role": "user",
            "content": [
                {"type": "text", "text": """
                    This image is a panorama from Google Street View.
                    From the image, extract the following information, in JSON format:
                    green: Percentage of the image that is green space (e.g. parks, gardens, trees, grass etc.). A number from 0-100.
                    environment: Classify the nature of the environment in this image. Built up/green/residential/shops/cafes?. A string.
                    water: If you see any streams/ponds/rivers/ocean in the image, estimate the distance to the water in meters. A number. If there is no water, return 0.
                    obscured: Proportion of view obscured by buildings (how much of total line of sight is blocked by buildings in close proximity). A number from 0-100.
                    people: the number of people you see in the image
                    cars: the number of cars you see in the image
                    bikes: the number of bikes you see in the image

                    Do not include comments in your JSON response. Only respond with the JSON object. Make sure the JSON is valid.
                """},
                {"type": "image"},
            ]
        }
    ]
    input_text = processor.apply_chat_template(messages, add_generation_prompt=True)
    inputs = processor(
        image,
        input_text,
        add_special_tokens=False,
        return_tensors="pt"
    ).to(model.device)

    for retry in range(3):
        output = model.generate(**inputs, max_new_tokens=5000)
        result = processor.decode(output[0])
        result = result[result.rindex("<|end_header_id|>") + len("<|end_header_id|>"):].strip().replace("<|eot_id|>", "")
        #print("Output:")
        try:
            result = json.loads(result)
            #pprint(result)
            row = row._asdict()
            row.update(result)
            results.append(row)
            print("\n")
            break
        except json.JSONDecodeError:
            print(f"Unable to parse: {result}")

results = pd.DataFrame(results)
results.to_csv("LLM_results.csv", index=False)
results

	Index	pid	n	time	anxiousness	latitude	longitude	geometry	panoid	panolat	panolon	panodate
0	13	P20006	1	2023-06-03T02:45:55Z	3	-36.892203	174.740125	POINT (174.7401253 -36.89220256)	AF1QipPv0qhRSftIHefRk_j2bH78qeRkOMi_8renBk2B	-36.891827	174.740463	2015-01-03
1	15	P20009	2	2023-05-17T04:54:48Z	3	-36.923191	174.748620	POINT (174.7486203 -36.92319093)	AF1QipNBp6ZTo1nm7EvX5wssSbZlHV7VaBaHqPsp3zre	-36.923299	174.744895	2016-03-30
9	24	P20021	1	2023-06-03T03:55:41Z	1	-38.140714	176.251862	POINT (176.2518616 -38.14071376)	AF1QipNiSoDDA2omwMtXrIq76eZTz6u7JXaIMz2lY2HN	-38.140783	176.251412	2017-04-01
10	26	P20021	3	2023-06-05T21:49:46Z	3	-36.894889	174.742775	POINT (174.7427751 -36.89488899)	AF1QipP403MtkkXU3Xsf07nh0U0n0oGiZni4MZoO2mnD	-36.896569	174.743840	2022-12-05
12	28	P20022	1	2023-04-25T02:58:09Z	3	-36.913364	174.729617	POINT (174.7296168 -36.91336394)	AF1QipPcsqbdyRALZRYd_9SvX3O-yqcWyFO9dVUBXvNz	-36.913070	174.727857	2016-07-29
...	...	...	...	...	...	...	...	...	...	...	...	...
456	575	P20304	2	2024-01-24T21:01:50Z	0	-36.922586	174.719866	POINT (174.7198659 -36.92258555)	AF1QipPKcFXiyarq--Lz0yNktLmcEaL-H3eGCF6DWmUi	-36.923321	174.719057	2022-02-11
457	584	P20314	2	2023-09-29T07:25:56Z	0	-36.899923	174.815464	POINT (174.8154642 -36.89992279)	AF1QipPeF_kdk6m_6Xml3oxzrkyIga1lBDodkWZ1Re78	-36.900272	174.815460	2015-06-08
458	597	P20318	1	2023-09-29T05:14:26Z	1	-36.893269	174.744269	POINT (174.7442694 -36.89326884)	AF1QipM9tZLFQoQbz4xJuy8LPLwCOf5xIyJhU8DxRdIv	-36.892409	174.745253	2018-06-11
459	598	P20318	2	2023-09-29T23:47:24Z	1	-36.854193	174.729131	POINT (174.7291308 -36.85419324)	AF1QipNqZ74PoOatlFqQRJGfJikvb6x_topGXM9XdRy3	-36.855219	174.731047	2016-11-20
460	599	P20318	3	2023-09-30T05:29:26Z	1	-36.893262	174.744341	POINT (174.744341 -36.89326166)	AF1QipODQrKPpYlQauG8zcZSenz1FU3Tsg_ApD0UGvx4	-36.892527	174.745399	2018-06-11